Cold Email Sending Infrastructure for Agencies: How the AGP Machine Runs It

Cold email sending infrastructure for agencies is the layer that gets your emails into inboxes: dedicated sending domains (never your main domain), SPF, DKIM and DMARC on every one, a pool of mailboxes, and rules that spread volume under daily limits. Done right, it lets an agency send at real scale while its main domain stays untouched. This post shows how that layer runs inside Artemis, the system behind the AGP growth machine.

Nerdy Joe is owned by Agency Growth Partners (AGP), so this is our own system and we are not neutral about it. We will show you exactly how it works, give you criteria you can apply to any provider or a DIY build, and say who it is not a fit for.

If you want the generic, tool-by-tool setup steps (buying domains, adding DNS records, picking a sending platform), read our cold email infrastructure setup guide first. This post is about operating that infrastructure as part of a full engine, day after day, at agency volume.

Where sending infrastructure sits in the machine

AGP runs a five-stage sequence for agencies: Advise, Build, Fill, Close and Partner. Sending infrastructure belongs to Build. It is the plumbing that every later stage depends on.

The stage before it is data. By the time a contact reaches a mailbox, it has already been enriched and verified, which we cover in verified lead lists for agencies. Every address is verified before it is sent to, so bounces do not start dragging down a new domain.

The stage after it is warmup and placement. New mailboxes do not go straight to full volume. They are warmed first and monitored daily, covered in mailbox warmup and inbox placement.

Here is how the infrastructure layer looks inside Artemis:

What Google and Yahoo require from senders

Before any design choice, know the rules the big mailbox providers publish. These are the floor, not the goal.

Google's sender guidelines set requirements for all senders and a stricter tier for anyone sending more than 5,000 messages a day to Gmail accounts. Yahoo publishes a matching set of all-sender and bulk sender requirements.

Google also asks senders to keep the spam rate below 0.10% and never let it reach 0.30%. It requires TLS for transmitting email and recommends 2048-bit DKIM keys where supported (1024 bits is the minimum).

The 5,000 threshold matters less than it looks. An agency running real outbound across many mailboxes can cross it without noticing, and you never want to find out from a delivery drop. Artemis treats every sending domain as if the bulk rules already apply: SPF, DKIM and DMARC on every one.

Rule 1: The main domain never sends cold email

Your main domain is where client invoices, proposals, contracts and support replies live. If cold email damages its reputation, all of that starts landing in spam. That is a business risk, not a marketing metric.

So inside Artemis, the client's main domain is never used for cold email. Outreach goes out from dedicated sending domains that exist only for outbound. If one of them gets tired, it can be rested or replaced while the main domain keeps working.

Yahoo makes the same underlying point: each IP and DKIM domain carries its own reputation, and it recommends separating mail types by IP or DKIM domain. Separate domains are how you keep cold outreach from sharing a reputation with your client email.

What to check in any setup

• Does any cold sequence send from the main domain, even "just a few"? It should not.
• Who owns the sending domains, and what happens to them if you end the engagement?
• Is each sending domain clearly tied back to the brand, so a prospect who looks it up is not confused?

Rule 2: Authenticate every domain and check every mailbox

SPF says which servers may send for a domain. DKIM signs each message so the receiver can confirm it was not altered. DMARC tells receivers what to do when a message fails, and it is the record that ties the visible From address to SPF or DKIM.

In Artemis, SPF, DKIM and DMARC are set on every sending domain. Then SPF and DMARC are checked again per mailbox, and each mailbox gets its own health view. Records drift: someone edits DNS, a provider changes, a mailbox gets added under the wrong domain. Checking per mailbox is how that gets caught before it costs placement.

A simple test for any provider: ask them to show you the authentication status of one specific mailbox, today. If they can only show you a domain-level setup from launch week, nobody is watching drift.

Rule 3: Mix mailbox providers on purpose

Artemis connects mailboxes on Google Workspace, Microsoft 365 or any SMTP/IMAP account. That flexibility matters for agencies for three practical reasons.

1. Existing accounts. Some clients already run on one provider and want to stay there. The engine adapts instead of forcing a migration.
2. Visibility by provider. Prospects use Gmail, Outlook and Yahoo inboxes. Sending from more than one provider, and measuring placement at Google, Microsoft and Yahoo, shows you where a problem actually is.
3. No single point of failure. If one provider tightens something, the whole engine does not stall.

What does not change across providers: every mailbox is authenticated, warmed, capped and monitored the same way.

Rule 4: Spread volume across mailboxes under daily limits

This is where agency infrastructure differs from a founder sending from a single inbox. Volume is a pool problem. Artemis spreads sends across mailboxes, keeps each one under its daily limit, and supports up to 500K emails a month per engine.

On AGP's sample dashboard, a mailbox shows a daily send count against its limit, for example 23 of 50. That is how the limit is meant to work: visible per mailbox, every day, not buried in a settings screen.

Here is an illustrative way to size a pool. This is arithmetic, not an AGP recommendation, and real caps depend on mailbox age and health.

The point of the table: hitting volume by raising per-mailbox caps is the wrong lever. You add mailboxes and domains, keep caps conservative, and leave headroom so a slipping mailbox can drop volume without missing the monthly target. For what that 50,000 a month turns into downstream, see agency outbound math.

Rule 5: Send inside windows, with spacing

Artemis sends inside sending windows and puts spacing between sends. Two things happen when you do this right.

First, emails land during the prospect's business day, when a real person is likely to read them. Second, each mailbox sends at a human pace instead of firing a batch at 9:00 a.m. Google advises senders to avoid sudden volume spikes and to increase volume slowly over time, and spacing is the day-level version of the same advice.

A quick checklist for windows:

• Window set to the prospect's time zone, not yours
• Spacing between sends from the same mailbox
• Weekends and holidays handled deliberately
• Sequences stop on reply, so nobody gets a follow-up after answering

Rule 6: Drop volume automatically when a mailbox slips

Every mailbox eventually has a bad week. The difference between a healthy engine and a burned one is what happens next.

In Artemis, volume drops automatically when a mailbox starts to slip, meaning it starts landing in spam. Nobody has to notice a chart and remember to change a setting. The mailbox sends less, recovers, and the rest of the pool carries the load.

AGP's sample dashboard shows what this looks like in practice: a panel of 48 mailboxes with 41 healthy, 6 warming and 1 resting at a health score of 71 and 84% placement. That is a sample, not an average or a promise. The useful part is the shape: most of the pool working, a few coming online, and a weak mailbox pulled back instead of pushed harder.

This is the mid-point of how AGP's growth machine works: data feeds infrastructure, infrastructure protects placement, placement makes personalization worth writing.

Who runs this day to day

At AGP, each client engine has a dedicated growth engineer who builds, runs and tunes it: lists, copy, deliverability and replies. Infrastructure is not a one-time setup they hand off. It is something they watch daily in the same live dashboard the client logs into.

That matters because most infrastructure failures are slow. A DNS record changes, one mailbox drifts, spam placement creeps up at one provider. Someone has to be looking.

How to judge any sending setup (AGP, another provider or DIY)

Use these questions whether you are hiring an agency, buying software or building it yourself. If you are comparing outside teams, our guide to what a cold email agency does covers the wider service.

Who this is not a fit for

Be honest with yourself about scale. If you only need a handful of conversations a month, one or two warmed mailboxes and careful manual sending may be enough. A full engine is overkill.

It is also not a fit if your offer is not ready. Infrastructure delivers emails; it does not make a weak offer land. AGP's services page puts it plainly: most agency pipelines fail at the offer, not the list. Fix that first.

And if you want to own every DNS record and mailbox yourself, with your own team running it, that is a valid choice. AGP offers a Build and Release model for exactly that: AGP builds the engine and trains your team to run it.

Common infrastructure mistakes agencies make

1. Sending from the main domain to "test." One test campaign can hurt the domain your clients email.
2. Authenticating once and never checking again. Drift is the normal state of DNS over time.
3. Chasing volume with caps. Raising limits on tired mailboxes speeds up the decline.
4. Launching cold mailboxes. New mailboxes need warmup before real volume.
5. Watching opens instead of placement. Placement by provider tells you where mail lands.
6. Ignoring unsubscribes. Yahoo asks bulk senders to honor them within 2 days. Artemis honors them automatically.

Want this built for your agency?

If you want the sending layer, and the rest of the full growth machine, built and run on Artemis for your agency, book a call with David. You will get a plain answer on whether it fits your offer and volume.

Sources

• Agency Growth Partners
• AGP: Artemis
• AGP: Services
• AGP: Results
• AGP: About
• AGP: Custom installs
• Gmail Help: Email sender guidelines
• Yahoo Sender Hub: Sender best practices
• Book a call with David